Session Audio Handling & Voice Toggles¶
Aura implements a session-based audio processing loop. Voice commands for state management are only active within an established recording session.
Configuration¶
The session-internal behavior is controlled by:
ENABLE_WAKE_WORD = True/False (in config/settings.py)
Operational Logic¶
Unlike a persistent background listener, Aura’s STT engine (Vosk) only processes audio when a recording session has been triggered externally (e.g., via Hotkey).
Aura is your telescope 🔭 for the PC: Control from a distance!
The In-Session Toggle (“Teleskop”)¶
When ENABLE_WAKE_WORD is set to True:
Trigger: The user starts a session manually.
Toggle: Saying “Teleskop” during the session toggles between ACTIVE and SUSPENDED states.
Behavior: This allows the user to “pause” and “resume” text processing using voice commands without terminating the audio stream.
Privacy & Efficiency¶
When ENABLE_WAKE_WORD is set to False (Default):
STT Suppression: While in a suspended state, calls to
AcceptWaveformandPartialResultare completely skipped.Privacy: No audio data is analyzed unless the system is in an explicit active state.
Resource Management: CPU usage is minimized by bypassing neural network analysis during suspension.
Latency & Performance¶
Instant Resume: Because the
RawInputStreamremains open throughout the session, switching from SUSPENDED back to ACTIVE has 0ms additional latency.Loop Timing: The processing loop operates at a ~100ms interval (
q.get(timeout=0.1)), ensuring near-instant response times.